CN115524340A - Corrugated board printing defect detection equipment and detection method thereof - Google Patents

Corrugated board printing defect detection equipment and detection method thereof Download PDF

Info

Publication number
CN115524340A
CN115524340A CN202211309011.9A CN202211309011A CN115524340A CN 115524340 A CN115524340 A CN 115524340A CN 202211309011 A CN202211309011 A CN 202211309011A CN 115524340 A CN115524340 A CN 115524340A
Authority
CN
China
Prior art keywords
frame
pressing
corrugated
matching
belt
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211309011.9A
Other languages
Chinese (zh)
Inventor
罗富文
曾伟光
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong Tai Yi Machinery Co ltd
Original Assignee
Guangdong Tai Yi Machinery Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong Tai Yi Machinery Co ltd filed Critical Guangdong Tai Yi Machinery Co ltd
Priority to CN202211309011.9A priority Critical patent/CN115524340A/en
Publication of CN115524340A publication Critical patent/CN115524340A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84Systems specially adapted for particular applications
    • G01N21/88Investigating the presence of flaws or contamination
    • G01N21/8851Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • BPERFORMING OPERATIONS; TRANSPORTING
    • B41PRINTING; LINING MACHINES; TYPEWRITERS; STAMPS
    • B41FPRINTING MACHINES OR PRESSES
    • B41F33/00Indicating, counting, warning, control or safety devices
    • B41F33/0036Devices for scanning or checking the printed matter for quality control
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84Systems specially adapted for particular applications
    • G01N21/88Investigating the presence of flaws or contamination
    • G01N21/89Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles
    • G01N21/892Investigating the presence of flaws or contamination in moving material, e.g. running paper or textiles characterised by the flaw, defect or object feature examined
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84Systems specially adapted for particular applications
    • G01N21/88Investigating the presence of flaws or contamination
    • G01N21/8851Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G01N2021/8887Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques

Landscapes

  • Engineering & Computer Science (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Pathology (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Immunology (AREA)
  • General Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Signal Processing (AREA)
  • Textile Engineering (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

The invention relates to a corrugated carton printing defect visual detection device and a detection method thereof, wherein the corrugated carton printing defect detection device comprises a rack, a transmission mechanism, a belt pressing structure and a shooting structure, the transmission mechanism is arranged on the rack and used for transmitting a corrugated paperboard, the belt pressing structure is arranged above the transmission mechanism and used for pressing the corrugated paperboard against the transmission mechanism, and the shooting structure is used for shooting the surface of the corrugated paperboard. The visual inspection equipment for the printing defects of the corrugated case is suitable for flexible production, has a visual inspection technology for automatically processing ultrahigh-resolution images on line at a high speed and performing tasks such as defect extraction, color difference detection, grading color separation and the like, and develops high-speed on-line inspection based on an artificial intelligent visual mechanism.

Description

Corrugated board printing defect detection equipment and detection method thereof
Technical Field
The invention relates to the field of printing equipment, in particular to a corrugated board printing defect detection device and a detection method thereof.
Background
In the traditional industrial production, because the limitation of science and technology is still mainly to adopt a manual detection method to detect the printing defects of the corrugated case, the method not only has low speed and low efficiency for detecting the products, but also is easy to make mistakes in the detection process, thereby causing the inaccuracy of the detection result. In the current society, with the appearance and development of scientific technologies such as artificial intelligence, machine learning and the like, a printing defect detection technology based on a machine vision technology appears, and the efficiency of printing defect detection is greatly improved.
The high-net-speed printing equipment for the corrugated case has the remarkable characteristics of high-net-line high-resolution printing, high speed, large amplitude, complex pattern textures and the like, and has high requirements on printing quality and stability, however, due to the fact that rear-end visual detection equipment is generally lacked in the industry, a closed-loop mode from production to detection cannot be formed, when the defects of color difference, missing beating, boolean lines, printing defects caused by faults of a spray head, disordered images, interference caused by environmental factors and the like occur in digital printing of corrugated paper, the defects can not be detected in time and found and intervened, a large number of defective products are caused, the rear-end high-speed visual online detection equipment is continuously developed, the real-time online detection of the printing quality is realized, and the printing quality and the stability are guaranteed.
A DH-JP1020 large-sheet product inspection machine derived from Beijing Daheng images can detect printed packages with the width of 1020mm, and the detection speed reaches 11000 m/h. The printed product is fed to the paper feeding and positioned accurately and then conveyed to the air sucking roller stably, the detection system shoots the image of the printed product by using a high-definition camera and stores the image in a computer memory, a standard detection template is firstly manufactured, and then the detection processing system is used for setting the quantitative standard of the detection of the printed product as the standard of the detection.
The CCIES printing product full-page quality detection system provided by Jiangsu scientific electromechanical engineering company is provided with an advanced printing product full-page quality detection machine, the system adopts a color linear array CCD camera to carry out image scanning to obtain the real image information of the surface of the printing product, the scanning speed reaches 180mm/s, and the scanning area reaches 1200 x 900mm.
The SMT automatic detection scheme proposed by Shenzhen Rui Xin Automation Limited company can realize the defects of patch technologies such as dislocation, wrong elements, missing elements, multiple elements, polarity reverse sticking and the like of patch elements.
Most of the products need strict light source conditions or are customized and debugged on site according to the needs of a production line, the design flow is complicated, and the flexibility degree is low. The equipment has the advantages that the detection speed is high, only specific defects can be detected, the generalization capability is not strong, the coexistence of the detection speed and high-resolution and high-precision detection is difficult to ensure, and the detection precision and the detection speed are all required to be improved.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to develop a visual detection technology which is suitable for flexible production and has the functions of automatically processing ultrahigh-resolution images on line at a high speed, extracting and detecting defects and the like, and a high-speed on-line device for detecting printing defects of corrugated boards based on an artificial intelligence visual mechanism.
The invention is realized by the following technical scheme:
a corrugated board printing defect detection device comprises a frame, a transmission mechanism, a belt pressing structure and a shooting structure, wherein the transmission mechanism is arranged on the frame and used for transmitting corrugated boards, the belt pressing structure is arranged above the transmission mechanism and used for pressing the corrugated boards, the shooting structure is used for shooting the surfaces of the corrugated boards,
the pressing belt structure comprises a pressing belt mounting frame and a pressing belt assembly arranged on the pressing belt mounting frame, the pressing belt mounting frame comprises an arch frame and two guide frames which are arranged on the arch frame in parallel, the pressing belt assembly comprises fixing seats, pressing wheels, a pressing belt, a connecting shaft, a pressing belt motor mounting seat and a pressing belt driving motor, the fixing seats are two and are respectively connected to the bottom surfaces of the two guide frames, the front end and the rear end of each fixing seat are provided with the pressing wheels and are sleeved with the pressing belt, the connecting shaft is one and is connected between a group of opposite pressing wheels of the two fixing seats, the motor mounting seat is fixedly arranged on one of the fixing seats and is provided with the pressing belt driving motor, a driving shaft of the pressing belt driving motor is in transmission connection with the pressing wheels to drive the pressing wheels and the connecting shaft connected with the pressing wheels to rotate, and thus the pressing belts on the two pressing belt assemblies are driven to synchronously rotate;
the shooting structure comprises a camera assembly and a light source assembly, the camera assembly comprises a lifting frame, a camera placing frame and cameras, the lifting frame comprises stand columns and lifting rods arranged at the top ends of the stand columns in a lifting mode, the stand columns are oppositely arranged on two sides of the feeding end of the stand frame, the camera placing frame is connected to the top ends of the two lifting rods, and the cameras are arranged on the camera placing frame and used for shooting the corrugated boards conveyed to the top surface of the stand frame;
the light source subassembly includes light source support, pivot and light source group, and the light source support includes two installation poles, and two both sides that the installation pole was fixed in the frame are higher than the frame top surface, corresponding to looks frame below, the pivot both ends are rotationally connected on the top of two installation poles, and the light source group is installed on the pivot for illuminate corrugated container board surface and in order to supply to shoot clearly.
Furthermore, the machine frame comprises machine frame long edges, support sectional materials and printing gold bars, the machine frame long edges are arranged in parallel, two ends of the two machine frame long edges are correspondingly formed into two opening ends of the machine frame, one end is a feeding end, and the other end is a discharging end; the support section bars are vertically connected between the two long edges, the support section bars are arranged at equal intervals along the long edges of the rack, and the two sheet metal strips are respectively and vertically fixedly arranged at two ports of the rack.
Furthermore, transport mechanism is the multiunit, connects gradually and installs in the frame, falls into the multistage transport structure who is connected in the frame, and each transport mechanism includes servo motor, shaft coupling, pivot, action wheel, conveyer belt, driven shaft and follows the driving wheel, and servo motor fixes in the long limit outside of frame, and servo motor's output has the pivot through the coupling joint, and the pivot is parallel sheet metal strip installs the action wheel in the pivot, and the driven shaft setting is in the long limit of frame and parallel with the pivot, installs on the driven shaft from the driving wheel, and each conveyer belt cover is established at the driving wheel to corresponding follow driving wheel.
Further, the arch frame includes two montants and connects the horizontal pole between two montants perpendicularly, and two montants are respectively established on two frame length edges of frame relatively, and the horizontal pole is connected at the top of two montants, is on a parallel with hold in the palm the section bar, the leading truck includes jib, guide bar and down tube, and the jib is connected in the bottom of horizontal pole perpendicularly, and the guide bar is connected in the bottom of jib perpendicularly, and the guide bar with the long limit of frame is parallel, and the down tube both ends are connected on jib and guide bar for the connection of the three-rod body is right triangle and connects.
Furthermore, the fixing seats are L-shaped plates and are provided with notches which are correspondingly surrounded, the directions of the notches of the two fixing seats are opposite, and pinch rollers are arranged at the front end and the rear end in the notches of each fixing seat and sleeved with press belts.
Further, the camera subassembly includes worm wheel lead screw lift and grab handle hand wheel, worm wheel lead screw lift installs and is used for controlling the lift height of lifter relative stand on the top of stand, and two worm wheel lead screw lifts are connected through a synchronizing bar that the level set up, and grab handle hand wheel is installed to one of them worm wheel lead screw lift, drives two lifter synchronous lifting through rocking handle hand wheel.
Furthermore, there are four lower margins in the bottom of the rack for stabilizing the rack, four cross shafts are arranged at the upper ends of the lower margins in a reinforcing manner, and four trundles are arranged at two ends of each cross shaft.
In addition, the invention provides a detection method of the corrugated board printing defect detection equipment, which uses the template picture and the picture shot by a binocular camera to carry out feature matching so as to carry out image detection, and the detection method comprises the following steps:
firstly, inputting a template picture S and a picture T obtained by shooting into a CNN network to respectively extract a coarse-grained feature map
Figure BDA0003907170190000041
Figure BDA0003907170190000042
And fine grain feature map
Figure BDA0003907170190000043
Expanding the obtained coarse-grained characteristic diagram, performing position coding by using a sin function, coding position information into a specific vector, and
Figure BDA0003907170190000044
and
Figure BDA0003907170190000045
fusing the expanded features to obtain fused features containing position information and depth feature information, and extracting the fused information between feature points in the same image and the two images by using a Transformer module with a self-attention layer and a cross-attention layerVisual association information of characteristic points among the characteristic points is obtained to obtain matching characteristics with specificity
Figure BDA0003907170190000046
And
Figure BDA0003907170190000047
matching the obtained features
Figure BDA0003907170190000048
And
Figure BDA0003907170190000049
using a micro-matchable layer for matching to obtain the positions i, j, by using features
Figure BDA00039071701900000410
And
Figure BDA00039071701900000411
calculating a scoring matrix Ms, and completing the processing by using a dual-softmax method to obtain a matching probability P c
Figure BDA00039071701900000412
P c (i,j)=softmax(Ms(i,·)) j softmax(Ms(·:j)) i
Where <, > is the inner product, the coefficient k is the temperature parameter, the result of the partial matrix Ms in softmax is made more pronounced by setting k to 0.1;
based on confidence matrix P c Confidence levels higher than θ may be selected c Further executing a nearest neighbor algorithm MNN to obtain a coarse-grained matching prediction M c (ii) a The coarse-grained match prediction is expressed as:
Figure BDA0003907170190000051
matching coarse-grained feature mapsThe resulting position
Figure BDA0003907170190000052
Mapping to fine-grained feature map, first mapping at fine-grained
Figure BDA0003907170190000053
And
Figure BDA0003907170190000054
to obtain the corresponding position
Figure BDA0003907170190000055
Based on the position, a pair of local windows with the size of s multiplied by s is cut out and input into a Transformer module with a self-attention layer and a cross-attention layer to obtain a local feature map
Figure BDA0003907170190000056
And with
Figure BDA0003907170190000057
Will be provided with
Figure BDA0003907170190000058
Central vector of and
Figure BDA0003907170190000059
calculating and correlating all vectors to generate a correlation heat map, obtaining a high-precision matching position by calculating expectation of probability distribution, and obtaining an integral matching item after repeating;
and detecting whether the corrugated case has printing defects or not through the matching item.
The wide-width rack is constructed, nine conveyor belts are vertically paved in the rack, the nine conveyor belts are respectively installed on nine driving wheels, and the driving wheels are driven by the servo motor so as to drive the conveyor belts to realize high-speed automatic conveying; in the detection process, based on the problem of difficulty in high-speed large-width real-time online detection, the device is characterized in that a belt pressing frame is arranged in the middle section of a conveying belt, a row of high-brightness illumination light sources are hung behind the belt pressing frame and used for ensuring that patterns can be shot clearly under the condition of insufficient brightness, two high-resolution cameras are fixed on a camera support above the belt pressing frame and used for shooting printing patterns of a corrugated carton under high-speed transmission, the shot patterns are input into a double-GPU host, and one-station detection equipment for ultrahigh-resolution image processing, high-precision image flaw extraction, complex texture detection and the like is realized through cooperation of high-resolution image detection based on a pixel level, a high-resolution image target extraction method based on a batch parallel processing mode and artificial intelligence and a pattern image color difference detection algorithm, so that automatic and intelligent industrial visual detection is realized.
Drawings
Fig. 1 is a perspective view of a corrugated cardboard printing defect detecting apparatus according to the present invention;
fig. 2 is a front view of the corrugated box defect visual inspection apparatus shown in fig. 1;
fig. 3 is a schematic view of a frame of the visual defect inspection equipment for corrugated cases shown in fig. 1;
FIG. 4 is a schematic view of a belt mounting rack of the visual defect inspection apparatus for corrugated cases shown in FIG. 1;
FIG. 5 is a diagrammatic view of a camera assembly of the visual defect inspection apparatus for corrugated containers shown in FIG. 1;
fig. 6 is a schematic view of a light source assembly of the visual defect inspection apparatus for corrugated cases shown in fig. 1.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1 and fig. 2, a printing defect detecting apparatus for corrugated cardboard according to a preferred embodiment of the present invention includes a frame, a conveying mechanism disposed on the frame for conveying the corrugated cardboard, a belt pressing structure disposed above the conveying mechanism for pressing the corrugated cardboard against the conveying mechanism, and a shooting structure for taking a picture of the surface of the corrugated cardboard.
Referring to fig. 3, the rack includes two long sides 2, a support section bar 1 and a printing plate bar 3, the two long sides 2 are arranged in parallel, two ends of the two long sides 2 are correspondingly formed as two open ends of the rack, one end is a feeding end, and the other end is a discharging end; support section bar 1 and connect perpendicularly between two long limits 2, and a plurality of support section bars 1 set up along frame long limit 2 equidistant, and the panel beating strip 3 is two, sets firmly respectively perpendicularly at frame both ends mouth.
The conveying mechanisms can be multiple groups and are sequentially installed on the rack in a connected mode, the rack is divided into multiple sections for conveying, so that corrugated boards can be conveyed in a butt joint mode through adjacent conveying mechanisms, in the embodiment, the conveying mechanisms are arranged into two groups, a front conveying section and a rear conveying section are formed on the rack, each conveying mechanism comprises a servo motor 5, a coupler 6, a rotating shaft 7, a driving wheel 8, a conveying belt 9, a driven shaft 10 and a driven wheel 11, the servo motors 5 are fixed to the outer side of the long edge 2 of the rack, the output ends of the servo motors 5 are connected with the rotating shafts 7 through the couplers 6, the rotating shafts 7 are parallel to the sheet metal strips 3, the driving wheels 8 are installed on the rotating shafts 7, the driven shafts 10 are arranged in the middle of the rack and are parallel to the rotating shafts 7, the driven wheels 11 are installed on the driven shafts 10, each conveying belt 9 is sleeved on the corresponding driven wheels 11, the driving wheels 8 are parallel to the corresponding driven wheels 10, in length of one section of the conveying belt is arranged between the driving wheels 8 and the corresponding driven wheels 10, in the embodiment, the two servo motors 5 of the two groups of conveying mechanisms are respectively fixed to the end, which is fixedly arranged on the rack and are arranged on the outer side of the middle of the rack, specifically, and two ends of the two groups of the rack are arranged in the driven shafts adjacent to the two ends of the rack.
Referring to fig. 4, the press belt structure includes a press belt mounting frame and a press belt assembly installed on the press belt mounting frame, the press belt mounting frame includes an arch 12 and a guide frame 18 installed on the arch 12, the arch 12 includes two vertical bars and a cross bar vertically connected between the two vertical bars, the two vertical bars are respectively and oppositely erected on the long sides 2 of the two racks of the racks, and the cross bar is connected to the tops of the two vertical bars and is parallel to the support profile 1. Leading truck 18 is two, installs on the horizontal pole parallelly for the installation is pressed the area subassembly, and leading truck 18 includes jib, guide bar and down tube, and the jib is connected perpendicularly in the bottom of horizontal pole, and the guide bar is connected perpendicularly in the bottom of jib, and the guide bar with the long limit 2 of frame is parallel, and the down tube both ends are connected on jib and guide bar for the connection of the three-pole body is right triangle and connects, with the stability of being connected of consolidating guide bar and jib.
Press the area subassembly including connecting block 19, fixing base 21, pinch roller 22, press the area 23, link 24, press area motor mount pad 25, and press area driving motor 26, connecting block 19 is two, connect the guide bar bottom surface at two leading framves 18 correspondingly through bolt and nut, fixing base 21 is two, connect respectively in two guide bar bottom surfaces, fixing base 21 becomes the L template, have the notch that the correspondence encloses, two fixing base 21's notch direction sets up relatively, pinch roller 22 installs in fixing base 21's notch, the front and back end sets up the pinch roller and overlaps and press area 23 in each fixing base 21's notch, it presses area 23 rotation to rotate the area to drive through pinch roller 22. The connecting shaft 24 is one and is connected between one group of pressing wheels 22 opposite to the two fixed seats 21, and the pressing wheel 22 on one fixed seat 21 rotates, so that the connecting shaft 24 correspondingly drives the pressing wheel 22 on the other fixed seat 21 to rotate, and the two pressing belts 23 rotate synchronously; the motor mounting seat 25 is fixedly mounted on one of the fixing seats 21 and is used for mounting the press belt driving motor 26, and a driving shaft of the press belt driving motor 26 is connected with the pressing wheel 22 and correspondingly drives the pressing wheel 22 to rotate, so that the press belt 23 is driven to rotate. In this embodiment, the installation seat 25 of the belt pressing motor is fixedly installed at one end of one of the fixed seats 21 having the connecting shaft 24, and the driving shaft of the belt pressing driving motor 26 correspondingly rotates the pressing wheel 22 connected with the connecting shaft 24 on the fixed seat 21 and drives the connecting shaft 24 to rotate together.
Further, one end of the cross bar of the arch frame 12 is vertically provided with a warning lamp 27.
Referring to fig. 5, the photographing structure includes a camera assembly and a light source assembly, the camera assembly includes a lifting frame 13, two worm screw lifters 31, two handle handwheels 32, a camera mount 33, a camera base 34 and a camera 35, the lifting frame 13 is arched and includes a vertical column and two lifting rods arranged at the top end of the vertical column in a lifting manner, the two vertical columns are oppositely arranged at two sides of the feeding end of the frame, the worm screw lifters 31 are arranged at the top end of the vertical column to control the lifting height of the lifting rods relative to the vertical column, the two worm screw lifters 31 are connected through a horizontal synchronization rod, one worm screw lifter 31 is provided with the handle handwheel 32, the two lifting rods are driven to be lifted synchronously through the rocking handle handwheel 32, two ends of the camera mount 33 are connected to the top ends of the two lifting rods, the camera mount 33 is driven to be lifted through the lifting rods, the camera base 34 is provided with a plurality of camera bases and arranged on the camera mount 33 for mounting the camera 35, and the camera 35 adopts a high-resolution camera facing the top surface of the frame and is used for photographing and conveying paper boards.
Referring to fig. 6, the light source assembly is used for providing brightness to the paper board for the camera 35 to take a picture clearly. The light source assembly comprises a light source support 14, a pivot 28 and a light source group 29, wherein the light source support 14 comprises two mounting rods, the two mounting rods are fixed on two sides of the rack through screws, are higher than the top surface of the rack and correspond to the lower part of the camera rack 13 and behind the press belt structure, two ends of the pivot 28 are rotatably connected to the top ends of the two mounting rods, and the light source group 29 is mounted on the pivot 28 to illuminate the conveyed paper board.
Furthermore, there are four lower feet 15 in the frame bottom and are used for stabilizing the frame, four cross shafts 16 are reinforced at the upper ends of the lower feet, and four casters 17 are installed at the two ends of each cross shaft.
In the embodiment of the invention, a wide-range rack is constructed, nine conveyor belts are vertically paved in the rack, the nine conveyor belts are respectively arranged on nine driving wheels, and the servo motor 5 drives the driving wheels 8 to drive the conveyor belts to realize high-speed automatic conveying; in the detection process, based on the problem of difficult high-speed large-width real-time online detection, the corrugated board printing defect detection equipment is provided with a belt pressing structure at the middle section of a conveying belt, a row of high-brightness illumination light sources are hung behind the belt pressing structure and used for ensuring that patterns can be clearly shot under the condition of insufficient brightness, two high-resolution cameras 34 are fixed on a camera mounting 33 above an arch frame 12, the high-resolution cameras 34 are used for shooting the printed patterns of a corrugated board box under high-speed transmission, the shot patterns are input into a double GPU host 50, and the ultrahigh-resolution image processing, the high-resolution image target extraction method based on a batch parallel processing mode, the artificial intelligent complex texture and pattern image color difference detection algorithm are cooperated to realize ultrahigh-resolution image processing, high-precision image flaw extraction, complex texture and pattern image detection and other one-station detection equipment, so that the automatic and intelligent industrial visual detection is realized.
The invention provides a detection method of corrugated board printing defect detection equipment, which uses a template picture and a double-camera shooting picture to perform feature matching so as to perform image detection.
Firstly, inputting a template picture S and a picture T obtained by shooting into a CNN network to respectively extract a coarse-grained characteristic graph
Figure BDA0003907170190000091
Figure BDA0003907170190000092
And fine grain feature map
Figure BDA0003907170190000093
Expanding the obtained coarse-grained characteristic diagram, and performing position coding by using a sin function;
encoding the position information into a specific vector, and
Figure BDA0003907170190000094
and
Figure BDA0003907170190000095
fusing the expanded features to obtain fused features containing position information and depth feature information, extracting visual association information of the fused information between feature points in the same image and feature points between the two images by using a Transformer module with a self-attention layer and a cross-attention layer to obtain specific matched feature
Figure BDA0003907170190000096
And
Figure BDA0003907170190000097
matching the obtained features
Figure BDA0003907170190000101
And
Figure BDA0003907170190000102
using a micro-matchable layer for matching to obtain the positions i, j, by using features
Figure BDA0003907170190000103
And with
Figure BDA0003907170190000104
Calculating a scoring matrix Ms, and completing the processing by using a dual-softmax method to obtain a matching probability P c
Figure BDA0003907170190000105
P c (i,j)=softmax(Ms(i,·)) j ·softmax(Ms(·,j)) i
Wherein<·,·>(i.e. the
Figure BDA0003907170190000106
) For inner products, the coefficient k is a temperature parameter, and the result of the partial matrix Ms in softmax is made more significant by setting k to 0.1.
Based on confidence matrix P c Confidence levels higher than θ may be selected c Further executing a nearest neighbor algorithm MNN to obtain a coarse-grained matching prediction M c (ii) a The coarse-grained match prediction is expressed as:
Figure BDA0003907170190000107
position obtained by matching coarse-grained characteristic diagram with result
Figure BDA0003907170190000108
Mapping to fine-grained feature map, first mapping at fine-grained level
Figure BDA0003907170190000109
And
Figure BDA00039071701900001010
get the corresponding position
Figure BDA00039071701900001011
Based on the position, a pair of local windows with the size of s multiplied by s is cut out and input into a Transformer module with a self-attention layer and a cross-attention layer to obtain a local feature map
Figure BDA00039071701900001012
And
Figure BDA00039071701900001013
will be provided with
Figure BDA00039071701900001014
Central vector of and
Figure BDA00039071701900001015
calculating all vectors to be correlated, generating a correlation heat map, calculating expectation of probability distribution to obtain a high-precision matching position, and repeating to obtain an integral matching item;
and detecting whether the corrugated case has printing defects or not through the matching items.
In summary, the present invention solves the following problems, aiming at the defects of the prior art schemes and the key common problem that visual detection in the same type of industries is urgently needed to break through:
(1) The high-speed large-width real-time online detection is difficult: the corrugated case has large width, the defect detection is required to be carried out through a high-resolution image, and the processing data amount is large; the high-speed digital printing machine has extremely high printing speed and is a great challenge for real-time online detection.
(2) Complex pattern flaws and color difference detection and identification are difficult: the corrugated paper digital printing machine can print high-resolution images, has fine and complex pattern textures, and is low in processing speed and precision when the traditional template matching mode is used for flaw extraction, color difference detection, grading color separation and the like.
(3) Flexible detection is difficult: the traditional template matching defect detection and color difference detection method at present needs to prepare a standard template in advance, and needs to change the matching template frequently when detecting various frequently-changed printing patterns, so that the adaptability is poor. While the efficiency cannot be guaranteed, a large amount of manpower is also spent.
The above description is only for the preferred embodiment of the present invention and is not intended to limit the scope of the present invention, and all equivalent structural changes made by using the contents of the present specification and the drawings, or any other related technical fields, are equally encompassed.

Claims (8)

1. The utility model provides a corrugated container board printing defect check out test set, includes frame and transport mechanism, and transport mechanism sets up and is used for conveying corrugated container board in the frame, its characterized in that: the device also comprises a press belt structure and a shooting structure, wherein the press belt structure is arranged above the conveying mechanism and used for pressing the corrugated board, the shooting structure is used for shooting the surface of the corrugated board,
the pressing belt structure comprises a pressing belt mounting frame and a pressing belt assembly arranged on the pressing belt mounting frame, the pressing belt mounting frame comprises an arch frame (12) and two guide frames (18) which are arranged on the arch frame (12) in parallel, the pressing belt assembly comprises two fixing seats (21), pressing wheels (22), pressing belts (23), a connecting shaft (24), a pressing belt motor mounting seat (25) and a pressing belt driving motor (26), the two fixing seats (21) are respectively connected to the bottom surfaces of the two guide frames (18), the front end and the rear end of each fixing seat (21) are provided with the pressing wheels and sleeved with the pressing belts (23), the connecting shaft (24) is one and is connected between a group of opposite pressing wheels (22) of the two fixing seats (21), the motor mounting seat (25) is fixedly arranged on one of the fixing seats (21) and provided with the pressing belt driving motor (26), a driving shaft of the pressing belt driving motor (26) is connected with the pressing wheels (22) in a transmission manner, the pressing wheels (22) and the connecting shaft (24) connected with the pressing wheels (22) are driven to rotate, so as to drive the pressing belts (23) on the two pressing belt assemblies to synchronously rotate;
the shooting structure comprises a camera assembly and a light source assembly, the camera assembly comprises a lifting frame (13), camera supports (33) and cameras (35), the lifting frame (13) comprises stand columns and lifting rods which are arranged at the top ends of the stand columns in a lifting mode, the stand columns are oppositely arranged on two sides of the feeding end of the stand frame, the camera supports (33) are connected to the top ends of the two lifting rods, and the cameras (35) are arranged on the camera supports (33) and used for shooting corrugated boards conveyed by the top surface of the stand frame;
the light source subassembly includes light source support (14), pivot (28) and light source group (29), and light source support (14) include two installation poles, and two installation poles are fixed in the both sides in the frame, are higher than the frame top surface, corresponding to camera frame (13) below, pivot (28) both ends are rotationally connected on the top of two installation poles, and light source group (29) are installed on pivot (28) for illuminate corrugated container board surface and make the picture clear.
2. A corrugated board printing defect detecting apparatus according to claim 1, characterized in that: the frame comprises frame long edges (2), support sectional materials (1) and printing gold bars (3), wherein the two frame long edges (2) are arranged in parallel, two ends of the two frame long edges (2) are correspondingly formed into two open ends of the frame, one end of each open end is a feeding end, and the other end of each open end is a discharging end; the support section bars (1) are vertically connected between the two long edges (2), the support section bars (1) are arranged at equal intervals along the long edges (2) of the frame, and the two sheet metal strips (3) are respectively and vertically fixed at two ports of the frame.
3. A corrugated board printing defect detecting apparatus according to claim 2, characterized in that: transport mechanism is the multiunit, connects gradually and installs in the frame, falls into the multistage transport structure that is connected in the frame, and each transport mechanism includes servo motor (5), shaft coupling (6), pivot (7), action wheel (8), conveyer belt (9), driven shaft (10) and follow driving wheel (11), and servo motor (5) are fixed in the long limit of frame (2) outside, and the output of servo motor (5) is connected with pivot (7) through shaft coupling (6), and pivot (7) are parallel sheet metal panel (3), install action wheel (8) on pivot (7), and driven shaft (10) set up between the long limit of frame (2) and are parallel with pivot (7), install on driven shaft (10) from driving wheel (11), and each conveyer belt (9) cover is established and is gone up from driving wheel (8) to corresponding follow driving wheel (11).
4. A corrugated board printing defect detecting apparatus according to claim 2, wherein: arch (12) include two montants and connect the horizontal pole between two montants perpendicularly, and two montants are established respectively relatively on two frame long limits (2) of frame, and the horizontal pole is connected at the top of two montants, is on a parallel with hold in the palm section bar (1), and leading truck (18) include jib, guide bar and down tube, and the jib is connected perpendicularly in the bottom of horizontal pole, and the guide bar is connected perpendicularly in the bottom of jib, and the guide bar with the long limit of frame (2) are parallel, and the down tube both ends are connected on jib and guide bar for the connection of the three-bar body is right angled triangle and connects.
5. A corrugated board printing defect detecting apparatus according to claim 1, characterized in that: the fixing seats (21) are L-shaped plates and are provided with notches which are correspondingly surrounded, the directions of the notches of the two fixing seats (21) are opposite, and the front end and the rear end in the notch of each fixing seat (21) are provided with press wheels and are sleeved with press belts (23).
6. A corrugated board printing defect detecting apparatus according to claim 2, characterized in that: the camera component comprises a worm wheel screw rod lifter (31) and a handle hand wheel (32), the worm wheel screw rod lifter (31) is installed on the top end of the stand column and used for controlling the lifting height of the lifting rod relative to the stand column, the two worm wheel screw rod lifters (31) are connected through a synchronizing rod arranged horizontally, the handle hand wheel (32) is installed on one worm wheel screw rod lifter (31), and the two lifting rods are driven to lift synchronously through rocking the handle hand wheel (32).
7. A corrugated board printing defect detecting apparatus according to claim 2, wherein: there are four lower margins (15) in the frame bottom to be used for firm frame, lower margin (15) upper end adds four inherent cross axles (16), four truckles (17) are installed at cross axle (16) both ends.
8. A method of inspecting a corrugated board printing defect inspection apparatus according to any one of claims 1 to 7, wherein: the method uses a template picture and a double-camera shooting picture to carry out feature matching so as to carry out image detection, and the detection method comprises the following steps:
firstly, inputting a template picture S and a picture T obtained by shooting into a CNN network to respectively extract a coarse-grained characteristic graph
Figure FDA0003907170180000031
Figure FDA0003907170180000032
And fine grain feature map
Figure FDA0003907170180000033
Expanding the obtained coarse-grained characteristic diagram, and performing position coding by using a sin function;
encoding the position information into a specific vector, and
Figure FDA0003907170180000034
and
Figure FDA0003907170180000035
fusing the expanded features to obtain fused features containing position information and depth feature information, extracting visual association information of the fused information between feature points in the same image and feature points between the two images by using a Transformer module with a self-attention layer and a cross-attention layer to obtain specific matched feature
Figure FDA0003907170180000036
And with
Figure FDA0003907170180000037
Will getTo match features
Figure FDA0003907170180000038
And
Figure FDA0003907170180000039
using a micro-matchable layer for matching to obtain the positions i, j, by using features
Figure FDA00039071701800000310
And with
Figure FDA00039071701800000311
Calculating a scoring matrix Ms, and completing the processing by using a dual-softmax method to obtain a matching probability P c
Figure FDA00039071701800000312
P c (i,j)=softmax(Ms(i,·)) j ·softmax(Ms(·,j)) i
Where <, > is the inner product and the coefficient k is the temperature parameter, the result of the partial matrix Ms in softmax is made more significant by setting k to 0.1;
based on confidence matrix P c Confidence levels higher than θ may be selected c Further executing a nearest neighbor algorithm MNN to obtain a coarse-grained matching prediction M c (ii) a The coarse-grained match prediction is expressed as:
Figure FDA0003907170180000041
position obtained by matching coarse-grained characteristic diagram with result
Figure FDA0003907170180000042
Mapping to fine-grained feature map, first mapping at fine-grained
Figure FDA0003907170180000043
And
Figure FDA0003907170180000044
get the corresponding position
Figure FDA0003907170180000045
Based on the position, a pair of local windows with the size of s multiplied by s is cut out and input into a Transformer module with a self-attention layer and a cross-attention layer to obtain a local feature map
Figure FDA0003907170180000046
And with
Figure FDA0003907170180000047
Will be provided with
Figure FDA0003907170180000048
Central vector of (2) and
Figure FDA0003907170180000049
calculating and correlating all vectors to generate a correlation heat map, obtaining a high-precision matching position by calculating expectation of probability distribution, and obtaining an integral matching item after repeating;
and detecting whether the corrugated case has printing defects or not through the matching items.
CN202211309011.9A 2022-10-25 2022-10-25 Corrugated board printing defect detection equipment and detection method thereof Pending CN115524340A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211309011.9A CN115524340A (en) 2022-10-25 2022-10-25 Corrugated board printing defect detection equipment and detection method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211309011.9A CN115524340A (en) 2022-10-25 2022-10-25 Corrugated board printing defect detection equipment and detection method thereof

Publications (1)

Publication Number Publication Date
CN115524340A true CN115524340A (en) 2022-12-27

Family

ID=84703840

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211309011.9A Pending CN115524340A (en) 2022-10-25 2022-10-25 Corrugated board printing defect detection equipment and detection method thereof

Country Status (1)

Country Link
CN (1) CN115524340A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220056165A1 (en) * 2019-05-09 2022-02-24 South China Agriculture University Supported three-center catalyst and preparation method and application
TWI846410B (en) * 2023-03-30 2024-06-21 高頡建材股份有限公司 Surfboard testing equipment

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220056165A1 (en) * 2019-05-09 2022-02-24 South China Agriculture University Supported three-center catalyst and preparation method and application
TWI846410B (en) * 2023-03-30 2024-06-21 高頡建材股份有限公司 Surfboard testing equipment

Similar Documents

Publication Publication Date Title
CN115524340A (en) Corrugated board printing defect detection equipment and detection method thereof
JP7055212B2 (en) How to inspect masks online
JP2012159489A (en) Optical inspection apparatus
CN1212523C (en) Inspection system for inspecting discrete wiring patterns formed on continuous substrate sheet of flexible material
CN111524184B (en) Intelligent unstacking method and unstacking system based on 3D vision
CN105973898A (en) Detection device of intelligent-terminal touch-screen protection cover plate
WO2013106904A1 (en) Automatic online defective elements-checking-and-removing device for flexographic printing presses
CN104722496A (en) Online defect detecting and sorting device of photovoltaic cell panel
CN212301356U (en) Wheel hub welding seam visual detection device
CN113030095A (en) Polaroid appearance defect detecting system
CN111729860A (en) Online automatic detection system and method for appearance defects of washing machine
KR102577055B1 (en) Error checking device for variable printing substrates
US20130181396A1 (en) Automatic Online Defective Elements-Checking-and-Removing Device for Flexographic Printing Presses
CN219122063U (en) Online quality detection device for printed matter
CN212422544U (en) Printing machine with detection device
CN211989743U (en) Cylindrical workpiece detection platform
CN213779923U (en) Shoes clothes broken pin detection device based on artificial intelligence technique
CN2908575Y (en) Fabric flat net printing machine of automatic register
CN213813399U (en) Detection equipment based on panel turnover mechanism
CN114120173A (en) Cigarette posture adjusting and error correcting system based on visual image recognition technology of high-speed moving object
CN206838541U (en) A kind of bamboo and wood bar separator
CN206235574U (en) Temperature controller testing equipment and temperature controller detection production line
Liu Machine Learning-based Automatic Defect Classification in Real Manufacturing Settings
CN112474401A (en) Visual inspection classification method and system for ceramic resistor substrate size
CN218707543U (en) Printing quality intellectual detection system equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination